AI in Workforce Management: Past, Present, and Future
Explore how AI is currently infused in modern WFM technologies and what AI innovations are on the horizon.
By:
I recently had the pleasure of presenting at the CRMXchange Virtual WFO Conference in a session entitled, AI & WFM: Busting Myths and Setting the Record Straight.
My colleague and co-presenter, Trudy Cannon, Senior Director, Go-to-Market Strategy, and I had a wonderful discussion not only about myths and rumors we are hearing about Artificial Intelligence (AI) and Workforce Management (WFM), but also about how AI is currently infused in modern WFM technologies and what AI innovations are on the horizon.
In my role as Verint’s Vice President of Product Strategy for WFM, I’m continually researching ways AI can enhance workforce planning. Here’s a summary of my thoughts on AI and WFM—past, present and future, and what contact center workforce planners can do to prepare for current and future AI innovations in WFM.
AI in Workforce Management Today
Workforce Management solutions have been around for decades, even though many organizations still rely on spreadsheets and homegrown tools. Modern WFM technologies have actually been using AI for some time now, in the form of:
- Algorithms and machine learning models
- Scheduling engines and automated processes
- Natural language, intelligent virtual assistant (IVA) interfaces
Algorithms and Machine Learning Models
WFM has been using algorithms, such as Erlang C, Erlang A, and Poisson distribution, and machine learning to help generate forecasts for years. However, these algorithms and machine learning models were only available to data scientists—people who knew how to code.
There has been a lot of science and a lot of growth in the algorithm and machine learning model space, adding hundreds of different models that deal with time series problems.
Today those models and algorithms have been democratized. You can now grab a model off the shelf and test that model in your organization.
Scheduling engines and automated processes
Scheduling engines and the automated processes included in those also use AI, such as simulated annealing, genetic algorithms and constraint satisfaction optimization.
AI makes these much easier to leverage. For example, you can go onto Amazon, use their AWS capabilities and build your own simulated annealing algorithm. That doesn’t mean you can suddenly build a successful WFM practice or WFM solution.
There’s a big gap between slapping together off-the-shelf algorithms and forecasting models and actually creating highly accurate forecasts and the best staffing plans you’ve ever had. It just doesn’t end up working like that. You need to understand the nuances of your business operations, handle time variability, seasonal patterns, and how to tune these models to your specific contact center environment.
Natural Language IVA Interfaces
Another use of AI in WFM is natural language or IVA interfaces.
Using large language models (LLMs), we’re starting to have conversations with the AI. This is changing how we work in the forecasting, the analytics, and upfront in the actual IVR—which has replaced previous systems such as the classic dual-tone multi-frequency or touch-tone (DTMF) systems.
Current Innovations in AI in Workforce Management
The next big innovation that’s currently being explored is retrieval augmented generation (RAG), which is an AI reference library that the AI can check in real time before answering a question.
RAG helps fix the issue with early LLMs that would, frankly, “hallucinate”—make up facts when they didn’t know/find something.
Open AI started building out these big data sets of contextual information. You can have all your data, whether it be in Salesforce, Confluence, ServiceNow, or an automated WFM solution, in one giant data source that AI can interact with. With RAG, AI doesn’t just rely on one-time training, but rather it pulls live data from right now.
Let’s say you had a knowledge article that you updated. That goes into the RAG, and it becomes the next best action/source. Instead of an AI guessing about PTO policies, RAG will retrieve your company’s actual handbook and cite the correct policy, thus reducing hallucinations.
What’s on the Horizon for AI in Workforce Management
All the rage right now is the hype around Agentic AI—the move from rules-based workflows to true automation that not only answers questions but can also take action on behalf of people while keeping humans in the loop. However, we’re very early in that agentic worker phase.
WFM Agentic Workers
We are working on creating the best WFM agentic workers to partner with your operations teams. So, for WFM, that means autonomous schedule optimization.
It means proactive identification and resolution of things such as compliance issues, intelligent break scheduling based on real time and incoming volume, and automated shift-bidding processes where you don’t have to be so entrenched in the shift-bidding process.
In the future, AI will be able to do a lot of the conflict resolution and make most of those decisions for you—and be able to give you the layout of what it’s choosing and why.
In the next few years, agentic workers will be able to get the forecast ready for you. AI can make you a great baseline forecast using one of these world-class algorithms we discussed (or choosing the best algorithm for each channel, work type and scenario).
However, the accuracy of that forecast will be incomplete (myth #1 in the webinar I mentioned earlier), because the AI has zero context for the things that could potentially happen—things the Workforce Planner would know.
Workforce Planners Still Reign
There's a deep business context that we as humans have and decision-making criteria that we manage, which are really, really important to forecasting, planning and scheduling. AI in WFM forecasting, can’t—and shouldn’t—be allowed to operate by itself.
The good news is two-fold.
- This means the workforce planner’s job will not be replaced (myth #2 in the webinar I mentioned earlier), and
- The AI will be an immense time saver for WFM Planners. It can do all the not so pleasurable tasks of data gathering, analysis, running Python scripts, selecting and testing algorithms, etc. (Check out the story of Juan, a Workforce Analyst and how his role was reimagined with AI-powered WFM.)
In the next three to four years, you’re going to see things such as forecasting workflows getting more automated and less complicated. For example, I foresee AI will:
- Make many of the calculations you currently do within Excel automatically
- Incorporate all work types, channels and corresponding handle times
- Factor in model attrition
- Automatically add variables from external data sources and calculate their impact.
All of this will give you more statistically valid and accurate forecasts.
What WFM Planners Can Do To Be Prepared for the Future
Moving forward, workforce planners need to become really good at giving instructions. We’re moving to a world where I see a lot more communication with chatbots to do a lot of work inside of WFM. We’re not there yet, but it’s coming.
The ability to communicate clearly and precisely with large language models will be critical—whether it be ChatGPT, Claude, Grok or whatever LLM you’re going to use—to be able to have a conversation with our WFM solution.
And the use of feedback loops with those large language models is a skill all its own, and it takes a lot of practice—both to get insights and also to give instructions.
Suddenly, when you string together a set of instructions that both an agentic worker, a large language model, and RAG can use, AI can go and do tasks and processes for you. A warning though: there’s a level of fluency that’s actually very technical that you will need to learn as you go through the process.
Want to learn more about AI and WFM? Watch the webinar: AI and WFM: Busting Myths and Setting the Record Straight.